Conflicts of Interests Among Shareholders: The Case of Corporate Acquisitions *

Transcription

1 Conflicts of Interests Among Shareholders: The Case of Corporate Acquisitions * Jarrad Harford University of Washington Dirk Jenter Massachusetts Institute of Technology and NBER Kai Li University of British Columbia First Draft: June 2006 This Version: June 2007 Abstract: We identify important conflicts of interests among shareholders and examine their effects on corporate decisions. When a firm is considering an action that affects other firms in its shareholders portfolios, shareholders with heterogeneous portfolios will disagree about whether to proceed. This effect is measurable and potentially large in the case of corporate acquisitions, where bidder shareholders with holdings in the target want management to maximize a weighted average of both firms equity values. Empirically, we show that such cross-holdings are large for a significant group of institutional shareholders in the average acquisition and for a majority of institutional shareholders in a significant number of deals. We find that managers consider cross-holdings when identifying potential targets and in structuring deals. Further, targets selected on cross-holdings will have lower synergies on average, which offsets the tendency for cross-holdings to lead to more aggressive bidding. Overall, we conclude that conflicts of interests among shareholders are sizeable and, at least in the case of acquisitions, affect managerial decisions. Keywords: cross-holdings, shareholder heterogeneity, target selection, mergers and acquisitions, toehold, synergies, agency JEL classification: G30, G34 * We thank Nittai Bergman, Murray Carlson, Adlai Fisher, Ron Giammarino, Jon Karpoff, Alan Kraus, Kalina Manova, Wayne Mikkelson, Pablo Moran, Hernan Ortiz-Molina, Jon Reuter, Frederik Schlingemann, Jeremy Stein, and Ralph Walkling for useful comments and discussions, and seminar participants at Bentley College, Boston College, Concordia, MIT finance lunch, Rutgers University, Simon Fraser University, University of British Columbia, University of Texas at Austin, University of Toronto, and York University, as well as participants of the Pacific Northwest Finance Conference (Seattle), and the 13 th Mitsui Life Symposium on Value Creation at the University of Michigan (Ann Arbor). We acknowledge the financial support from the Social Sciences and Humanities Research Council of Canada. All remaining errors are our own. University of Washington Business School, Box , Seattle, WA , , jarrad@u.washington.edu. Sloan School of Management, Massachusetts Institute of Technology, E52-446, Cambridge, MA 02142, , djenter@mit.edu. Sauder School of Business, University of British Columbia, 2053 Main Mall, Vancouver, BC V6T 1Z2, , kai.li@sauder.ubc.ca.

2 Conflicts of Interests Among Shareholders: The Case of Corporate Acquisitions Abstract We identify important conflicts of interests among shareholders and examine their effects on corporate decisions. When a firm is considering an action that affects other firms in its shareholders portfolios, shareholders with heterogeneous portfolios will disagree about whether to proceed. This effect is measurable and potentially large in the case of corporate acquisitions, where bidder shareholders with holdings in the target want management to maximize a weighted average of both firms equity values. Empirically, we show that such cross-holdings are large for a significant group of institutional shareholders in the average acquisition and for a majority of institutional shareholders in a significant number of deals. We find that managers consider cross-holdings when identifying potential targets and in structuring deals. Further, targets selected on cross-holdings will have lower synergies on average, which offsets the tendency for cross-holdings to lead to more aggressive bidding. Overall, we conclude that conflicts of interests among shareholders are sizeable and, at least in the case of acquisitions, affect managerial decisions. Keywords: cross-holdings, shareholder heterogeneity, target selection, mergers and acquisitions, toehold, synergies, agency JEL classification: G30, G34

3 Much of the corporate finance literature takes for granted the assumption that (homogenous) shareholders want to maximize the value of their firm s equity. In reality, most shareholders are diversified and care about the effects of managerial actions on their portfolio values, not just on the value of the specific firm taking the action. When shareholders have different portfolios, they may disagree about which actions managers should take. In acquisitions, diversified shareholders who hold equity stakes in both a bidder and its target (cross-holdings) want bidder management to maximize a weighted average of the bidder s and the target s equity values. Concentrated shareholders with stakes in only the bidder want management to focus on the bidder s equity value, setting-up a conflict between shareholder groups. Managers must choose how to act in the presence of heterogeneous shareholders with different portfolios. They could ignore this heterogeneity and simply align themselves with the concentrated shareholders. Alternatively, they could take it into account in an attempt to maximize the wealth of most stockholders, and/or opportunistically use it to lower shareholder resistance to empire-building acquisitions. In this study, we first measure the extent of shareholder cross-holdings and then test for their impact on target choice and deal characteristics to determine whether managers incorporate cross-holdings into their acquisition decisions. The insight that cross-holdings by shareholders alter their preferences over takeover decisions is a special case of the general result that diversified shareholders prefer corporate policies which maximize their portfolio values to policies which narrowly maximize the values of individual firms (Hansen and Lott (1996), and Rubin (2006)). In general, however, externalities on other firms imposed by managers decisions are likely to be small or hard to estimate, so in a bounded-rationality framework, it makes sense for managers to focus on own firm value. In corporate acquisitions, one of the externalities is large and easily computed, and so the opportunity to correctly maximize the wealth of the majority (but not simultaneously all) of the shareholders is present. Thus, we use merger bids as our experimental setting and specifically 1

4 test the hypothesis that cross-holdings by bidder shareholders affect target selection and deal structure. We formally show that the preference of a bidder shareholder with cross-holdings in the target is captured by the ratio of her percentage stake in the target to her percentage stake in the bidder, and that this preference is equivalent to the preference of a value-maximizing bidder with a toehold in the target. In all models of takeover bidding with toeholds considered in the prior literature, toeholds make bidders more willing to pursue an acquisition and more willing to bid aggressively. Bidding firms that consider their shareholders cross-holdings are therefore expected to follow a more aggressive acquisition strategy because their shareholders will capture some of the premium paid. The net effect of cross-holdings on bidding strategies and bid outcomes is more ambiguous. Under the null hypothesis that cross-holdings have an effect on mergers, managers will choose targets that have high cross-holdings, high synergies, or both. Thus, among targets with high cross-holdings, some will also have high synergies, but many will not. Although the direct effect of cross-holdings would be to increase premiums and completion likelihoods, the fact that they may signal low synergies produces an indirect opposing effect. As a result, while cross-holdings should have an observable effect on target selection and bidder announcement returns, they have an ambiguous effect on premiums and deal completion. Bidder managers who want to take their shareholders cross-holdings into consideration face the challenge that different shareholders have different portfolios and hence different crossholdings in the target, with many bidder shareholders holding no shares in the target at all. The heterogeneity in investors cross-holdings leads to disagreements about the objective function for the bidding firm, and it is far from obvious which decision rule managers should adopt to determine their optimal acquisition strategy. We therefore use several measures of the distribution of shareholder preferences in the empirical analysis. An intuitively appealing solution to the aggregation problem is for managers to maximize a weighted average of their 2

5 shareholders preferences. This implies that a bidder shifts to a more aggressive acquisition strategy even if only 10 or 20 percent of its shareholders own shares in the target. Alternatively, management may follow a majority rule and simply adopt the preferences of its median shareholder. The median shareholder in our context is defined such that 50 percent of the bidder s shares are held by shareholders with smaller cross-holdings and 50 percent by shareholders with larger cross-holdings. We document the magnitude of investor cross-holdings in takeovers and the resulting conflicts of interests using information on the portfolios of large institutional investors and takeover data from 1984 to We show that the average bidder s median institutional shareholder 1 has a cross-holding of 5.3 percent in the target, with a median of 0.0 percent. The distribution of cross-holdings is strongly right-skewed both across bidders and across shareholders in a given bidder; in five percent of the bidders, the median institution has crossholdings of 40 percent or more in the target. In the average bidder, 20 percent of the institutional holdings are by investors with cross-holdings of more than 34 percent, and 10 percent of the institutional holdings are by investors with cross-holdings above 75 percent. Hence many institutional investors want bidder management to act as if the bidder had a large toehold in the target, which in turn implies substantial disagreements between shareholders with cross-holdings and those without. We compute the effect of these cross-holdings on the total abnormal returns experienced by institutional investors around the bids, including the effect of the bid on the target s stock price. We find that the median institution averages a positive 0.2 percent return improvement, or about 17 percent of the average bidder announcement return of 1.2 percent. We find that the presence of cross-holdings by bidder shareholders is important in determining the target. Target firms have significantly higher cross-holdings by bidder shareholders than do non-target control firms. Further, consistent with the conjecture that any 1 The median institutional shareholder is defined such that 50% of the institutional ownership in the bidder is by institutions with larger cross-holdings and 50% by institutions with smaller cross-holdings. This definition is similar to the definition of the median shareholder discussed before, with all percentiles re-defined relative to the total institutional ownership in the firm. 3

6 positive direct effect of cross-holdings on the aggressiveness of bidding strategies is mitigated by an indirect selection effect associating cross-holdings with low synergies, we find no effect of cross-holdings on bid premiums or deal completion probabilities. Cross-holdings are associated with lower post-acquisition operating performance, further supporting our conjecture that bidders trade off cross-holdings with synergies when selecting acquisition targets. Finally, we do not find evidence that managers are using cross-holdings to facilitate empire-building acquisitions. We conclude that the conflicts of interests between diversified and concentrated shareholders in mergers are pervasive enough to matter, and that managers pay attention to the cross-holdings of their diversified shareholders. In related work, Matvos and Ostrovsky (2006) argue that investor cross-holdings may explain the low and often negative returns to acquiring firms in takeovers. They argue that bidder shareholders with large cross-holdings do not mind overpaying for the target, and hence that cross-holdings help explain why shareholders fail to block apparently bad takeover deals. Different from our work, they do not analyze individual cross-holdings and instead add up bidder shareholders stakes in the target. This approach aggregates away the conflicts between shareholders with heterogeneous cross-holdings. Our analysis clarifies that shareholders preferences are determined by their relative holdings in the bidder and the target, and that only shareholders who own a larger percentage of the target than of the acquirer favor overpayment. Empirically, shareholders with cross-holdings above 100 percent are rare and hold on average only 4.1 percent of the bidders equity in the mergers in our data. Hence the influence of investors who favor overpayment is likely to be limited, and cross-holdings are unlikely to explain why bidders overpay. 2 2 Matvos and Ostrovsky (2006) also show that mutual funds that hold shares in both a bidder and its target are more likely to vote for mergers with negative announcement returns. This finding suggests that institutional investors do pay attention to the interactions between the firms in their portfolios. There has also been other work on shareholder conflicts. Eckbo and Verma (1994) identify shareholder conflicts over dividend policy due to heterogeneous tax rates, information asymmetries, and agency costs. Rice (2006) demonstrates that the importance of a shareholder group to price formation depends on the group s price elasticity of demand for shares, leading to a divergence between the effects of corporate decisions on market values and the effects of those decisions on shareholder 4

7 The plan of the paper is as follows. The next section draws on prior theoretical work on toeholds to formally analyze the effects of cross-holdings on firms objectives, and to develop our null hypothesis and some testable implications. Section II describes our data and the variable definitions used in the empirical analysis. Section III presents summary statistics on the size and prevalence of cross-holdings in actual takeovers and examines their effects on target selection, announcement returns, takeover premiums, and the probability of takeover success. The last section summarizes and concludes. I. Theoretical Background and Hypothesis Development This section examines how cross-holdings affect shareholders preferences over mergers and acquisitions, and how bidder management may react to the heterogeneous preferences of its shareholders. We first analyze shareholder preferences in the two canonical classes of takeover models analyzed in the prior literature, models of tender offers in the spirit of Grossman and Hart (1980) and models of competitive takeover bidding in the spirit of Fishman (1988). We then present our null hypothesis and its testable implications. I.A. Cross-holdings in a model of tender offers with free-riding This section describes shareholder preferences in a Grossman and Hart (1980)-type setting with dispersed target shareholders. To illustrate the idea, consider a model with two firms, a bidder and a target. The value of the target under incumbent management is V I. The bidder is able to create synergies worth S > 0 that are realized only if the bidder acquires control. To obtain control, the bidder needs to acquire 50 percent of the target shares in a conditional tender offer, which entails a transaction cost of k > 0. Assume that S > k so that the acquisition is socially efficient. All target shares are held by dispersed shareholders. welfare. Kraus and Rubin (2007) show that the conflict between diversified shareholders who want to maximize their portfolio values and managers who use capital budgeting rules to choose firm-value maximizing projects has important implications for executive compensation design. 5

8 Now consider the preferences of a shareholder who owns α B percent of the bidder and α T percent of the target. This shareholder wants the bidder to pursue an acquisition whenever her share of the associated takeover gains exceeds her share of the takeover costs: α T α T S > α B k k + S > 0 (1) α B The shareholder wants bidder management to put a weight of (α T / α B ) on target value and to effectively internalize (α T / α B ) percent of the takeover gains accruing to target shareholders. But the decision rule in (1) is simply the optimal rule for a value-maximizing bidding firm that owns a toehold of (α T / α B ) percent in the target (see for example, Shleifer and Vishny (1986), Hirshleifer and Titman (1990), and Chowdhry and Jegadeesh (1994)). Hence a bidder shareholder with relative cross-holdings of (α T / α B ) wants bidder management to act as if the bidder itself had a toehold of (α T / α B ). In the extreme, a shareholder who owns a higher percentage of the target than of the bidder wants management to act as if the bidder had a toehold larger than 100 percent in the target, which is to say that this shareholder would actually benefit from overpayment for the target. 3 I.B. Cross-holdings in a model of competitive takeover bidding This section examines shareholder preferences in models (for example, Fishman (1988), and Burkhart (1995)), in which two or more bidders compete for a target, and in which the Grossman-Hart free-rider problem plays no role. For simplicity, we assume that there are only two bidders, labeled 1 and 2, that both bidders are risk neutral, and that there are no bidding costs. Let V i denote the private value of the target to bidder i, and assume that both V 1 and V 2 exceed the value of the target firm under incumbent management. This ensures that the target will be taken over by one of the bidders in equilibrium, and that all target shares are acquired by 3 It may seem surprising that the shareholder s preference is independent of the values of her positions in the two firms, or equivalently, of the size of the bidder relative to the target. However, this result obtains because the impact of a larger relative bidder value on the shareholder s wealth (α B V B ) is exactly offset by the smaller effect on the bidder value due to overpaying for a relatively smaller target. 6

9 the winning bidder. Bidder 1 s and bidder 2 s strategies consist of choosing price ceilings C 1 and C 2, respectively, at which to withdraw from the auction, given that the rival is still actively bidding. Now consider the preferences of a shareholder who owns α B percent of bidder 1 and α T percent of the target. This shareholder receives α B percent of any takeover gains accruing to bidder 1 in case bidder 1 wins the auction, and the shareholder sells her α T percent target stake at the winning price to whichever bidder wins the auction. Hence her expected payoffs are given by: Pr(bidder 1 wins) Pr(bidder 1wins) V 1 E [ α ( V E( C bidder 1wins) ) + α E( C bidder 1wins) ] + Pr(bidder 2 wins) [ α C ] B 1 α T αt ( C bidder 1wins) + E( C bidder 1wins) + Pr(bidder 2 wins) C 2 2 α B T 2 2 T α B The payoff function puts a relative weight of one on the takeover gains to bidder 1 in case that bidder wins, and a relative weight of (α T / α B ) on the takeover price paid, independently of which bidder wins. But this is exactly the objective function of a bidding firm that owns a toehold of (α T / α B ) percent in the target and sells that toehold to the rival bidder in case the rival wins the auction (Burkhart (1995) and Singh (1998)). It follows that a bidder shareholder with relative cross-holdings of (α T / α B ) would see her expected wealth maximized if the bidder set its bidding strategy as if the bidder itself had a toehold of (α T / α B ) in the target. To summarize, we have shown that both in free-rider models based on Grossman and Hart (1980) and in models of competitive bidding based on Fishman (1988) and Burkhart (1995), a shareholder with stakes of α B percent in the bidder and α T percent in the target wants bidder management to act as if the bidder itself had a toehold of size (α T / α B ). This motivates our focus on the distribution of these (relative) cross-holdings among bidder shareholders in Section III. 1 (2) 1 7

10 I.C. Aggregation of shareholder preferences by bidder management The relative cross-holdings are by construction shareholder-specific and range from zero to values above 100 percent in most bidder-target combinations. Bidder management thus faces the question of how to aggregate the preferences of its shareholders into a decision rule for the overall firm. Even if we assume that bidder managers wish to maximize the overall welfare of their shareholders, it is far from obvious what acquisition strategy managers should adopt. With different cross-holdings by different shareholders, any policy has to favor the preferences of one subset of bidder shareholders over others. Theory provides little guidance as to how bidder management should aggregate its shareholders preferences, and what weight to attach to the target value in its objective function. Two intuitively appealing solutions to the aggregation problem are for managers to adopt the preferences of its median shareholder, or for managers to maximize a weighted average of shareholder preferences. Under the median-shareholder criterion, bidder management pursues all takeover deals which are supported by the owners of a majority of its shares. The median shareholder is defined such that 50 percent of bidder shares are held by investors with larger cross-holdings and 50 percent by investors with smaller cross-holdings. Since larger cross-holdings make a shareholder prefer a more aggressive acquisition strategy, the median shareholder defines the most aggressive strategy managers can implement while retaining the support of a majority of shareholders votes. The median-shareholder criterion by design ignores the preferences of most shareholders. Alternatively, bidder managers may try to maximize a weighted average of all their shareholders subjective preferences. While it is not obvious what weights management should use to aggregate shareholder preferences, most reasonable weighting schemes would imply that bidders shift to a more aggressive acquisition strategy even if only 10 or 20 percent of their shareholders 8

11 have equity stakes in a target. 4 In Section III, we document both the preferences of the median shareholder and of the subsets of bidder shareholders with the highest cross-holdings, and examine whether their preferences are reflected in acquisitions. The above aggregation schemes for shareholder preferences can only be relevant to realworld acquisitions if bidder management does in fact consider the portfolios held by its shareholders, or if diversified shareholders have the means to impose portfolio-value maximization on management. A priori, there are several reasons to expect that managers may ignore the cross-holdings their shareholders have in other firms. Large concentrated shareholders such as founding families or venture capitalists are unlikely to own any cross-holdings in target firms, but are likely to be active in corporate governance and to exert influence over managers. Further, the compensation of top executives is heavily biased towards stock options and restricted stock and thus ultimately depends on own-firm performance (Murphy (1999)). 5 Hence managers are given strong incentives to maximize the performance of their own firm, but few obvious incentives to care about any externalities they impose on other firms in their shareholders portfolios. The degree to which cross-holdings will matter thus depends on the ability of crossholding investors to influence bidder management. Shareholders with cross-holdings may be able to influence acquisition decisions through their representatives on boards of directors, voting at shareholders meetings, or private or public communication with the management team (Smith (1996), and Carleton, Nelson, and Weisbach (1998)). This kind of influence is most likely for 4 Drèze (1974) proposes to use shareholders ownership stakes as weights when aggregating heterogeneous shareholder preferences. Interestingly, applying this weighting scheme to relative cross-holdings amounts to simply adding up the bidder shareholders percentage stakes in the target, and hence gives the same aggregate preferences as if there were a perfect coalition among bidder shareholders. This degree of coordination between shareholders is not realistic and we opt not to use this aggregation scheme in the empirical analysis. We do, however, consider the possibility that the largest institutional shareholders may align their preferences through negotiated side-payments. 5 This raises the question of why managerial contracts are written this way to begin with. While beyond the scope of this paper, we offer the following considerations. First, incentives are endogenous and negotiated between managers and the board. If concentrated shareholders dominate corporate governance and boards, then incentive contracts will reflect their preferences. Second, moving away from firm-value maximization as the objective function in designing executive compensation schemes may have costs, e.g., create more opportunities for managers to maximize their own welfare at the expense of shareholders. 9

12 institutional investors and for blockholders, and for this reason our empirical analysis focuses on the cross-holdings of institutional investors with large stakes in the bidder. Whether bidder management takes its shareholders portfolios into account when making acquisition decisions is ultimately an empirical question which we attempt to answer in Section III. I.D. Hypothesis development and testable implications Our null hypothesis is that bidder managers take their shareholders cross-holdings into account in target selection and structuring deals, while our alternative hypothesis is that crossholdings have no effect on mergers. Assuming that the null hypothesis prevails, how would the bidder s acquisition strategy change? In Grossman and Hart (1980), a toehold bidder receives some of the takeover gains accruing to target shareholders, making it more likely that the bidder is able to cover the acquisition cost and hence more likely that the bid goes ahead. Crossholdings by bidder shareholders in a target should thus make an acquisition bid more likely. Thus, we expect: Testable Implication 1: There is a positive association between cross-holdings by a bidder s shareholders and, ceteris paribus, the likelihood that a firm will be targeted by that bidder. In models of competitive takeover bidding, a toehold makes it optimal for a bidder to bid more aggressively and to even raise its bid above its own valuation for the target. The more aggressive bidding strategy increases the bidder s expected capital gain on the toehold should a rival win and purchase the toehold. Hirshleifer and Titman (1990), and Chowdhry and Jegadeesh (1994) conclude that the probability of takeover success increases in the size of the bidder s toehold. Cross-holdings by bidder shareholders in a target should therefore motivate bidder management to bid more aggressively. Empirically, however, we only observe the net effect of cross-holdings on target selection and on bid aggressiveness. If firms select targets by trading off cross-holdings with synergies, then a substantial fraction of targets with high cross-holdings will have lower than average 10

13 synergies. Shleifer and Vishny (1986), and Hirshleifer and Titman (1990) show that toeholds are associated with lower bid premiums as toeholds allow acquisitions with lower value-added to proceed. To the extent that we are unable to perfectly measure synergies, this confounds our analysis. For example, a bid where a high premium is paid for a target with low synergies and high bidder shareholder cross-holdings may be indistinguishable from a bid where a normal premium is paid for a target with high synergies and low shareholder cross-holdings. Moreover, the benefits of more aggressive bidding accrue only to bidder shareholders with cross-holdings and not to the bidding firm itself. We predict: Testable Implication 2: The net effects of cross-holdings on premiums and completion rates are ambiguous, while their effect on bidder announcement returns is negative. Testable Implication 3: Ceteris paribus, there is a negative association between crossholdings and synergies. Under the null hypothesis, bidder managers may choose to take their shareholders crossholdings into account for the opportunistic purpose of reducing their resistance to empirebuilding acquisitions. If the motivation is management driven, then we would still expect to see an effect of cross-holdings in target selection, but not necessarily in bid characteristics. More importantly, we would expect to see this type of managerial behavior concentrated in poorly governed firms. Empirically, if agency problems drive the effect of cross-holdings on target selection, we expect: Testable Implication 4: There is a positive association between cross-holdings and poor governance in the bidding firm. II. Sample Formation and Variable Construction II.A. Sample selection We begin with all announced (both completed and cancelled) US mergers with announcement dates between January 1, 1984 and December 31, 2002 as identified from the 11

14 Mergers and Acquisitions database of the Securities Data Company (SDC). We identify all deals where both the bidder and target are public firms and the form of deal was coded as a merger, an acquisition of majority interest, or an acquisition of assets. After applying the above filters, we get 8,075 deals. The sample period is chosen because the information in SDC is less reliable before 1984 and we require data on post-merger performance. Next, we match our bidders and targets with Compustat and CRSP data, and we only retain an acquisition if the bidder owns less than 50 percent of the target prior to the bid and is seeking to own greater than 50 percent of the target. For completed deals, we require that the bidder owns more than 90 percent of the target after the deal completion. These filters yield 2,885 deals. Finally, we merge our acquisition data with the CDA/Spectrum 13F data on institutional shareholdings in the bidder and the target. 6 Ideally, we would also like to measure cross-holdings at the individual investor level. Using data on institutional investors adds a layer of intermediation between individual investors and firms, and makes it difficult to assess the extent to which the ultimate owners of the assets are diversified across firms. However, given the greater size of their stakes, the cross-holdings of institutional investors are more likely to be reflected in company policies than the cross-holdings of the individuals behind these institutions. When interpreting our results it is nevertheless useful to keep in mind that we do not observe the portfolios of non-institutional investors and therefore miss potentially significant cross-holdings by individual shareholders. In a similar vain, an institutional portfolio reported to the SEC may be an aggregate of multiple distinct portfolios managed by the institution, adding noise to our measure of cross-holdings. Our final sample has 2,815 merger deals where both the bidder and the target have data on institutional shareholdings in the quarter-end prior to the bid announcement. 6 A 1978 amendment to the Securities and Exchange Act of 1934 requires all institutions with greater than $100 million of equity securities under discretionary management to report every quarter using the SEC s form 13F all common-stock positions greater than 10,000 shares or $200,000. These reports are available in electronic form back to 1980 from CDA/Spectrum, a firm hired by the SEC to process the Form 13F filings. 12

15 II.B. Variable construction In Section I we show that bidder shareholders preferences are determined by their crossholdings, defined as their percentage ownership of the target divided by their percentage ownership of the bidder. Since different shareholders have different cross-holdings, the resulting shareholder preferences over acquisitions are heterogeneous. We therefore adopt three complementary approaches to describing the (cumulative) distribution of cross-holdings across shareholders. Given that we do not observe the portfolios of non-institutional investors, we focus most of our discussion on the cross-holdings of institutional investors and restrict our sample to bidders that have at least 20 percent of their equity owned by institutions. This restriction excludes firms in which institutional investors are unlikely to have any influence on management. The first approach to measuring shareholder preferences captures what percentage of the institutional holdings in the bidder is by investors with large cross-holdings, and what percentage by investors with small or no cross-holdings. Figure 1 illustrates the approach. Specifically, we order all institutional investors in the bidder by their cross-holdings, and then calculate the (marginal) cross-holding for the top 1, 5, 10, 20, and 50 percent of institutional shareholdings (labeled as the top-1%, top-5%, top-10%, top-20%, and median cross-holdings, respectively). For example, a top-10% cross-holding of 75 percent means that 10 percent of the institutional ownership in the bidder is controlled by investors with cross-holdings of at least 75 percent. Probably the most relevant measure of investor incentives in this set is the cross-holding of the median institutional shareholder in the bidder. The cross-holding of this investor, called the median cross-holding for short, determines the set of acquisition policies the bidder can pursue while maintaining the support of the majority of its institutional shareholders votes. To complement the analysis for institutional shareholders, we also calculate and report crossholdings for all shareholders of the bidder by making the conservative assumption that all noninstitutional investors have zero cross-holdings. 13

16 The above approach to measuring cross-holdings ignores the fact that bidder shareholders differ greatly in their ability to impose their preferences on bidder management. We expect bidder management to be most responsive to the preferences of its largest shareholders; hence our next two approaches look explicitly at the cross-holdings of blockholders, defined as institutional investors who own at least five percent of the bidder s shares. Specifically, we order the blockholders by their cross-holdings, and then report the (marginal) cross-holding for the top 5, 10, 20, 50, 75, and 100 percent of all shares owned by blockholders, in parallel to the calculations for all institutional investors described above. Finally, we consider the possibility that blockholders might coordinate their actions and negotiate side-payments with each other. With costless bargaining, the Coase Theorem (Coase (1960)) predicts that the blockholders should act in unison and support any acquisition that is profitable for all blockholders combined. The aggregate preferences of this coordinated group of investors are determined by their combined cross-holdings, given by the sum of their equity stakes in the target divided by the sum of their stakes in the bidder. In reality, coordination between blockholders may not be costless and side-payments are likely to be restricted by law. 7 Hence the blockholders combined cross-holdings should be interpreted as an upper bound on the cross-holding the blockholders may bring to bear on the bidding firm s decision process. III. Empirical Results In this section, we first present an overview of our sample with a focus on the magnitude of shareholder cross-holdings. Then, we formally test the null hypothesis that shareholder crossholdings do matter in target selection, bid characteristics, and outcomes in a multivariate regression framework. Finally, we present evidence on the indirect effect of cross-holdings on 7 The Securities and Exchange Commission passed new rules in 1992 allowing shareholders to directly communicate with each other (SEC (1992)). Thus, the costs of creating shareholder coalitions were substantially reduced. 14

17 bid characteristics and outcomes due to endogenous target selection and explore alternative explanations to our findings. III.A. Sample overview Table 1 presents descriptive statistics on the announced merger deals, the bidders, and the targets. In Panel A, we show that 75 percent of the announced deals are eventually completed. A quarter of the deals use only cash as the method of payment, and over 40 percent of the deals are pure stock swaps. Approximately 13 percent of the targets receive competing bids within one year. The sample is split evenly between diversifying and within industry deals. The mean relative deal size, defined as the ratio of the transaction value to the market value of the bidder, is 30 percent. Consistent with the prior literature, there are few direct toeholds, and more than 95 percent of the bidders in our sample have no toehold at all in their targets. On average, the bidder is offering a premium of 45 percent above the market price of the target as of four weeks before the bid. In Panel B, we show that the average three-day abnormal announcement period return (CAR3) for the bidder is 1.2 percent, while the average CAR3 for the target is close to 20 percent. Following Bradley, Desai, and Kim (1988), we also compute the dollar amount of the synergistic gains (= bidder CAR3 bidder market capitalization + target CAR3 (1 toehold) target market capitalization) as well as the percentage synergy gains (= synergy in dollars /(bidder market capitalization + (1 toehold) target market capitalization). Panel B shows that the average synergies amount to about $47 million in 2002 dollars, corresponding to an average percentage gain of about 2 percent. This implies that, once we account for the large positive announcement return to the target, mergers in our sample are on average welfare-improving. Panels C and D present summary statistics on the bidders and targets in our sample, respectively. Consistent with the relative size variable, the bidders are much larger than their targets in terms of both book and market values. The bidders and targets have similar levels of 15

18 leverage, while the bidders appear to enjoy faster sales growth, better operating performance, and higher stock returns in the year prior to the bid. Table 2 presents summary statistics of institutional shareholdings in the bidders and targets. On average, institutional investors own 46 percent of the equity of bidding firms and 32 percent of the equity of targets. Bidder institutions that also own shares in the target control 14 percent of all bidder shares, or 31 percent of the bidder shares owned by institutions. While target institutions that also own shares in the bidder control a similar 18 percent of all target shares, they own over half of the generally smaller institutional holdings in the target. III.B. The size and pervasiveness of cross-holdings Table 3 summarizes the empirical distribution of cross-holdings in the bidders and targets. In Part I of Panel A, we calculate cross-holdings for the top 1, 5, 10, 20, and 50 percent of the bidders institutional shareholders and limit our sample to bidders with total institutional ownership no less than 20 percent. This reduces our sample by only 16 percent to 2,356 observations. We find that the average cross-holding by the median institutional shareholder is 5.3 percent, with a median of 0.0 percent. The distribution of cross-holdings across bidders is strongly right-skewed; in five percent of the bidders, the median institution has cross-holdings of 40 percent or more, while in one percent of the bidders, the median cross-holding is at least 87 percent. Focusing on the largest cross-holdings, we find that the average top-10% cross-holding is above 75 percent, and that the average top-5% cross-holding is 140 percent. In Part II of Panel A, we calculate cross-holdings for the top 1, 5, 10, 20, and 50 percent of all bidder shareholders, assuming that the cross-holdings of non-institutional investors are zero. With this assumption, we find that the average median cross-holding is only 0.3 percent. This low number is a direct result of the assumed zero cross-holdings for non-institutional investors and of the fact that the median shareholder is an institution in less than half of the bidders in our sample. However, even under these assumptions, the average top-10% cross- 16

19 holding is 30 percent. We also compute the fraction of bidder shares whose holders own a higher percentage stake of the target than of the bidder. We find that bidder shareholders with crossholdings greater than one are rare and control on average only 4.1 percent of the bidder s equity and only 10.1 percent of the total institutional holdings in the bidder. This means that the influence of bidder shareholders who favor overpayment for the target is likely to be small. We conclude that there is a substantial number of institutional investors with large crossholdings in the average bidder, and that there is a significant set of mergers in which even the median institution owns a large stake in the target. At the same time, many institutional shareholders of the bidder have no cross-holdings in the target at all, implying significant conflicts of interests over the objective function of the bidder. The above analysis ignores the differences in the power of different shareholders to impose their preferences on bidder management. We therefore focus next on large shareholders. Table 4 presents summary statistics of cross-holdings held by blockholders, defined as institutions that own at least five percent of the shares of a firm. Part I of Panel A shows that the average cross-holdings by the median blockholder is 11 percent (median 0.0 percent), and that in five percent of the bidders, the median blockholder has a cross-holding of 82 percent or more in the target. Turning to blockholders with large cross-holdings, the average top-20% cross-holding exceeds 18 percent, and the average top-10% cross-holding is almost 20 percent. These numbers suggest that there is a significant number of institutional blockholders, the subset of institutional investors most likely to influence bidder management, with large cross-holdings in the target. In Part II of Panel A we examine what would happen if blockholders were to negotiate side-payments with one another and act as a group. The average combined cross-holding by all blockholders in the bidder is 12 percent. The distribution is again skewed, with the median of the combined cross-holding equal to zero. In ten percent of all deals, however, blockholders want management to act as if the bidder had a 46 percent toehold in the target, and in five percent of all deals as if the bidder had a 69 percent toehold in the target. For comparison, Betton, Eckbo, 17

20 and Thorburn (2005) analyze toeholds held directly by the acquirer in a comprehensive sample of twelve thousand bidders and find that only 11 percent of the bidders own any toehold in the target. In their sample, the average toehold size among bidders with a positive toehold is 21 percent, with a median of 17 percent. 8 Of course, target shareholders can have cross-holdings in the bidder as well, so Tables 3 and 4 present analogous information for target shareholders cross-holdings in the bidder. We find similar, but generally higher cross-holdings held by target shareholders in the bidder. This is likely a mechanical outcome driven by the relatively greater institutional interest in bidders observed in Table 2. III.C. Wealth effects of bid announcements Next we examine to what extent the wealth effects of takeovers on bidder shareholders change when we take their cross-holdings in the target into account. To address that question, we introduce the concept of a return improvement. First define the total abnormal announcement return for an institutional investor in a bidder as Total Abnormal Announcement Return = αbi VB + αti VT (3) where α is the percentage shareholding by institution i in either the bidder (subscript B) or in the target (subscript T) and V is the abnormal change in firm value, computed as the product of the bidder s or target s market value and the corresponding abnormal announcement period return. The return improvement is the difference between the total abnormal announcement return and the bidder abnormal announcement period return. This measure thus captures the improvement in return experienced by a given bidder shareholder due to her ownership stake in the target. Table 5 presents the return improvement for institutional investors in the bidder, identified by the size of their cross-holdings. Part I limits the sample to bidders with total 8 Similar, if somewhat higher, numbers have been documented before by, among others, Bradley, Desai, and Kim (1988), Jarrell and Poulsen (1989), Jennings and Mazzeo (1993), and Betton and Eckbo (2000). 18

21 institutional ownership no less than 20 percent. We find that the average return improvement for the median institutional shareholder is 0.2 percent, with a median of 0.0 percent. The distribution of return improvement is somewhat skewed; in five percent of the bidders, the median institution has a return improvement of 1.7 percent or better, while in one percent of the bidders, the median institution has a return improvement of 6.2 percent or better. Focusing on bidder institutions with large cross-holdings, we find that the average return improvement for the institution with the top- 10% cross-holdings is 2.5 percent; for the institution with the top-5% cross-holdings, it is 3.7 percent. As a reference point, the average bidder announcement return in our sample is 1.2 percent (see Table 1, Panel B). Thus, we conclude that, in general, the return improvement provided by cross-holdings is moderate, but that in some cases, it is large relative to the average announcement effect. Part II of Table 5 presents the return improvement for all shareholders, assuming individual shareholders have zero cross-holdings. There remain moderate improvements across bidder shareholders with non-trivial cross-holdings, although as expected the improvements are not as big as those in Panel A. III. D. The effect of cross-holdings on target selection In the next three sub-sections, we formally examine the null hypothesis that managers account for their shareholders cross-holdings when making decisions about acquisitions. The literature on toeholds has shown that the presence of a toehold increases the bidding likelihood, so we start our investigation by examining the role of cross-holdings in target selection. To examine the first testable implication of our null hypothesis, we compare the cross-holdings of bidder shareholders in the actually chosen targets to their cross-holdings in a sample of potential alternative targets that were not chosen. Our control sample of potential target firms is obtained by first identifying the set of all CRSP/Compustat firms that were neither targets nor bidders in our merger sample and that have 19

22 institutional shareholding data available from the CDA. This forms the base set of potential control firms. For each actual target in our sample, we select a control firm from its Fama-French (1997) industry that is closest in market capitalization (requiring that the difference does not exceed 25 percent of the sample firm market capitalization) at the fiscal year-end prior to the bid announcement. 9 Matching is done with replacement and only one control firm is matched to each sample firm. We are able to match 2,427 control firms to target firms in our sample. We refer to the matching firms as potential targets. In Table 6, Panel A, we present summary statistics for the potential targets, which can be compared to the corresponding statistics for the actual targets presented in Table 1, Panel D. There is no statistical difference between the actual and potential target samples across most firm characteristics, including the extent of institutional shareholding. This suggests that we have found valid control firms for the targets. The only significant difference between the potential and actual target firms is that the potential target firms tend to have higher market-to-book ratios. Table 6, Panel B presents the empirical distribution of cross-holdings by bidder shareholders in the potential targets. Comparing these figures to those in Panel A of Table 3, it is clear that the cross-holdings by bidder shareholders in the actual targets are larger than that in the potential targets (except where the cross-holdings for the median shareholder remain zero in both samples). Tests of differences in means and medians show that the cross-holdings by bidder shareholders in the chosen targets are significantly different from those in the control sample. An alternative explanation for the difference in cross-holdings are unobservables, that is, institutions may invest in firms that are similar along some dimension that we do not observe, and that same similarity may make those firms more likely to merge. While this explanation is plausible for within-industry acquisitions, it is unlikely to apply to diversifying acquisitions. It would have to be the case that institutions invest in firms across different industries based on some unobservable dimension that also makes these firms more likely to merge with each other. 9 Our main inferences remain unchanged if we impose different size matching criteria. 20